|
{ |
|
"best_metric": 0.7736489492761546, |
|
"best_model_checkpoint": "bart-base/checkpoint-4501", |
|
"epoch": 14.0, |
|
"eval_steps": 500, |
|
"global_step": 9002, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 0.78, |
|
"learning_rate": 4.871176775531364e-05, |
|
"loss": 1.2028, |
|
"step": 500 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"eval_accuracy": 0.7598760650658405, |
|
"eval_f1_macro": 0.6063644681337579, |
|
"eval_f1_micro": 0.7598760650658405, |
|
"eval_f1_weighted": 0.7489266744747745, |
|
"eval_loss": 0.8429557681083679, |
|
"eval_macro_fpr": 0.023226157303183535, |
|
"eval_macro_sensitivity": 0.6366904609623596, |
|
"eval_macro_specificity": 0.981694425945952, |
|
"eval_precision": 0.7600880594489849, |
|
"eval_precision_macro": 0.6003620126022547, |
|
"eval_recall": 0.7598760650658405, |
|
"eval_recall_macro": 0.6366904609623596, |
|
"eval_runtime": 43.5934, |
|
"eval_samples_per_second": 29.615, |
|
"eval_steps_per_second": 3.716, |
|
"eval_weighted_fpr": 0.02207348333808032, |
|
"eval_weighted_sensitivity": 0.7598760650658405, |
|
"eval_weighted_specificity": 0.9655403241234398, |
|
"step": 643 |
|
}, |
|
{ |
|
"epoch": 1.56, |
|
"learning_rate": 4.741575946086055e-05, |
|
"loss": 0.715, |
|
"step": 1000 |
|
}, |
|
{ |
|
"epoch": 2.0, |
|
"eval_accuracy": 0.7931835786212239, |
|
"eval_f1_macro": 0.7213629796675549, |
|
"eval_f1_micro": 0.793183578621224, |
|
"eval_f1_weighted": 0.7853477061503886, |
|
"eval_loss": 0.7331996560096741, |
|
"eval_macro_fpr": 0.018982515099395962, |
|
"eval_macro_sensitivity": 0.7320748785168393, |
|
"eval_macro_specificity": 0.9845140063777027, |
|
"eval_precision": 0.8019650926098448, |
|
"eval_precision_macro": 0.7386000214414028, |
|
"eval_recall": 0.7931835786212239, |
|
"eval_recall_macro": 0.7320748785168393, |
|
"eval_runtime": 77.2393, |
|
"eval_samples_per_second": 16.714, |
|
"eval_steps_per_second": 2.097, |
|
"eval_weighted_fpr": 0.018283914264192288, |
|
"eval_weighted_sensitivity": 0.7931835786212239, |
|
"eval_weighted_specificity": 0.9745265170443169, |
|
"step": 1286 |
|
}, |
|
{ |
|
"epoch": 2.33, |
|
"learning_rate": 4.611975116640747e-05, |
|
"loss": 0.578, |
|
"step": 1500 |
|
}, |
|
{ |
|
"epoch": 3.0, |
|
"eval_accuracy": 0.7939581719597212, |
|
"eval_f1_macro": 0.6998313050068915, |
|
"eval_f1_micro": 0.7939581719597211, |
|
"eval_f1_weighted": 0.79006368267302, |
|
"eval_loss": 0.8045273423194885, |
|
"eval_macro_fpr": 0.018545878882531782, |
|
"eval_macro_sensitivity": 0.7068834234702027, |
|
"eval_macro_specificity": 0.9847671550097226, |
|
"eval_precision": 0.8075005963568469, |
|
"eval_precision_macro": 0.7230632440844518, |
|
"eval_recall": 0.7939581719597212, |
|
"eval_recall_macro": 0.7068834234702027, |
|
"eval_runtime": 86.4099, |
|
"eval_samples_per_second": 14.94, |
|
"eval_steps_per_second": 1.875, |
|
"eval_weighted_fpr": 0.018199233716475097, |
|
"eval_weighted_sensitivity": 0.7939581719597212, |
|
"eval_weighted_specificity": 0.977549153186118, |
|
"step": 1929 |
|
}, |
|
{ |
|
"epoch": 3.11, |
|
"learning_rate": 4.482633488854329e-05, |
|
"loss": 0.481, |
|
"step": 2000 |
|
}, |
|
{ |
|
"epoch": 3.89, |
|
"learning_rate": 4.3530326594090203e-05, |
|
"loss": 0.3938, |
|
"step": 2500 |
|
}, |
|
{ |
|
"epoch": 4.0, |
|
"eval_accuracy": 0.8156467854376452, |
|
"eval_f1_macro": 0.7369315728111729, |
|
"eval_f1_micro": 0.8156467854376452, |
|
"eval_f1_weighted": 0.8105168035589256, |
|
"eval_loss": 0.8291025161743164, |
|
"eval_macro_fpr": 0.016938074954137138, |
|
"eval_macro_sensitivity": 0.7217605983180461, |
|
"eval_macro_specificity": 0.9857842035419169, |
|
"eval_precision": 0.8171220006778526, |
|
"eval_precision_macro": 0.7937034642937142, |
|
"eval_recall": 0.8156467854376452, |
|
"eval_recall_macro": 0.7217605983180461, |
|
"eval_runtime": 84.4366, |
|
"eval_samples_per_second": 15.29, |
|
"eval_steps_per_second": 1.919, |
|
"eval_weighted_fpr": 0.01588785046728972, |
|
"eval_weighted_sensitivity": 0.8156467854376452, |
|
"eval_weighted_specificity": 0.9711162676911109, |
|
"step": 2572 |
|
}, |
|
{ |
|
"epoch": 4.67, |
|
"learning_rate": 4.2236910316226025e-05, |
|
"loss": 0.3238, |
|
"step": 3000 |
|
}, |
|
{ |
|
"epoch": 5.0, |
|
"eval_accuracy": 0.7939581719597212, |
|
"eval_f1_macro": 0.7360579650857649, |
|
"eval_f1_micro": 0.7939581719597211, |
|
"eval_f1_weighted": 0.7995200606905476, |
|
"eval_loss": 0.8888681530952454, |
|
"eval_macro_fpr": 0.018767728188278533, |
|
"eval_macro_sensitivity": 0.7515467286591376, |
|
"eval_macro_specificity": 0.9846803211687324, |
|
"eval_precision": 0.8146370555192632, |
|
"eval_precision_macro": 0.7464262358644205, |
|
"eval_recall": 0.7939581719597212, |
|
"eval_recall_macro": 0.7515467286591376, |
|
"eval_runtime": 82.179, |
|
"eval_samples_per_second": 15.71, |
|
"eval_steps_per_second": 1.971, |
|
"eval_weighted_fpr": 0.018199233716475097, |
|
"eval_weighted_sensitivity": 0.7939581719597212, |
|
"eval_weighted_specificity": 0.9762466455712672, |
|
"step": 3215 |
|
}, |
|
{ |
|
"epoch": 5.44, |
|
"learning_rate": 4.094090202177294e-05, |
|
"loss": 0.246, |
|
"step": 3500 |
|
}, |
|
{ |
|
"epoch": 6.0, |
|
"eval_accuracy": 0.7955073586367157, |
|
"eval_f1_macro": 0.7362210043644567, |
|
"eval_f1_micro": 0.7955073586367158, |
|
"eval_f1_weighted": 0.7945977979999089, |
|
"eval_loss": 1.1628626585006714, |
|
"eval_macro_fpr": 0.01863187379355594, |
|
"eval_macro_sensitivity": 0.7599986341032186, |
|
"eval_macro_specificity": 0.9846932142915057, |
|
"eval_precision": 0.8066758597712774, |
|
"eval_precision_macro": 0.7483163388808308, |
|
"eval_recall": 0.7955073586367157, |
|
"eval_recall_macro": 0.7599986341032186, |
|
"eval_runtime": 82.1621, |
|
"eval_samples_per_second": 15.713, |
|
"eval_steps_per_second": 1.972, |
|
"eval_weighted_fpr": 0.018030323726266905, |
|
"eval_weighted_sensitivity": 0.7955073586367157, |
|
"eval_weighted_specificity": 0.9748908557358684, |
|
"step": 3858 |
|
}, |
|
{ |
|
"epoch": 6.22, |
|
"learning_rate": 3.964489372731986e-05, |
|
"loss": 0.228, |
|
"step": 4000 |
|
}, |
|
{ |
|
"epoch": 7.0, |
|
"learning_rate": 3.834888543286677e-05, |
|
"loss": 0.1791, |
|
"step": 4500 |
|
}, |
|
{ |
|
"epoch": 7.0, |
|
"eval_accuracy": 0.8179705654531371, |
|
"eval_f1_macro": 0.7736489492761546, |
|
"eval_f1_micro": 0.817970565453137, |
|
"eval_f1_weighted": 0.8146592022169725, |
|
"eval_loss": 1.1354283094406128, |
|
"eval_macro_fpr": 0.016454731140779742, |
|
"eval_macro_sensitivity": 0.7696557438486226, |
|
"eval_macro_specificity": 0.9861772746118799, |
|
"eval_precision": 0.8150792023791334, |
|
"eval_precision_macro": 0.7832166669456337, |
|
"eval_recall": 0.8179705654531371, |
|
"eval_recall_macro": 0.7696557438486226, |
|
"eval_runtime": 85.4548, |
|
"eval_samples_per_second": 15.107, |
|
"eval_steps_per_second": 1.896, |
|
"eval_weighted_fpr": 0.01564684732671949, |
|
"eval_weighted_sensitivity": 0.8179705654531371, |
|
"eval_weighted_specificity": 0.9746885537250626, |
|
"step": 4501 |
|
}, |
|
{ |
|
"epoch": 7.78, |
|
"learning_rate": 3.705546915500259e-05, |
|
"loss": 0.1305, |
|
"step": 5000 |
|
}, |
|
{ |
|
"epoch": 8.0, |
|
"eval_accuracy": 0.8109992254066615, |
|
"eval_f1_macro": 0.7368975448573482, |
|
"eval_f1_micro": 0.8109992254066615, |
|
"eval_f1_weighted": 0.808822910649843, |
|
"eval_loss": 1.282464861869812, |
|
"eval_macro_fpr": 0.016943107063520774, |
|
"eval_macro_sensitivity": 0.7488577570944955, |
|
"eval_macro_specificity": 0.9858309939135591, |
|
"eval_precision": 0.8147632227505013, |
|
"eval_precision_macro": 0.7422132710401315, |
|
"eval_recall": 0.8109992254066615, |
|
"eval_recall_macro": 0.7488577570944955, |
|
"eval_runtime": 87.6049, |
|
"eval_samples_per_second": 14.737, |
|
"eval_steps_per_second": 1.849, |
|
"eval_weighted_fpr": 0.016373641121997046, |
|
"eval_weighted_sensitivity": 0.8109992254066615, |
|
"eval_weighted_specificity": 0.9764656832967246, |
|
"step": 5144 |
|
}, |
|
{ |
|
"epoch": 8.55, |
|
"learning_rate": 3.575946086054951e-05, |
|
"loss": 0.0924, |
|
"step": 5500 |
|
}, |
|
{ |
|
"epoch": 9.0, |
|
"eval_accuracy": 0.8040278853601859, |
|
"eval_f1_macro": 0.7559721746710123, |
|
"eval_f1_micro": 0.8040278853601859, |
|
"eval_f1_weighted": 0.80145679310201, |
|
"eval_loss": 1.421708583831787, |
|
"eval_macro_fpr": 0.017802568540250287, |
|
"eval_macro_sensitivity": 0.7809236600744539, |
|
"eval_macro_specificity": 0.9853492039682624, |
|
"eval_precision": 0.8113877598541575, |
|
"eval_precision_macro": 0.7464644897074477, |
|
"eval_recall": 0.8040278853601859, |
|
"eval_recall_macro": 0.7809236600744539, |
|
"eval_runtime": 89.5999, |
|
"eval_samples_per_second": 14.408, |
|
"eval_steps_per_second": 1.808, |
|
"eval_weighted_fpr": 0.017111937774771727, |
|
"eval_weighted_sensitivity": 0.8040278853601859, |
|
"eval_weighted_specificity": 0.9762101741637506, |
|
"step": 5787 |
|
}, |
|
{ |
|
"epoch": 9.33, |
|
"learning_rate": 3.446345256609643e-05, |
|
"loss": 0.0953, |
|
"step": 6000 |
|
}, |
|
{ |
|
"epoch": 10.0, |
|
"eval_accuracy": 0.8024786986831913, |
|
"eval_f1_macro": 0.7661426626929498, |
|
"eval_f1_micro": 0.8024786986831914, |
|
"eval_f1_weighted": 0.8001228885030943, |
|
"eval_loss": 1.5552372932434082, |
|
"eval_macro_fpr": 0.018250858387380567, |
|
"eval_macro_sensitivity": 0.7822186507003115, |
|
"eval_macro_specificity": 0.9849087494473102, |
|
"eval_precision": 0.8055922813865947, |
|
"eval_precision_macro": 0.7702182213697708, |
|
"eval_recall": 0.8024786986831913, |
|
"eval_recall_macro": 0.7822186507003115, |
|
"eval_runtime": 85.2534, |
|
"eval_samples_per_second": 15.143, |
|
"eval_steps_per_second": 1.9, |
|
"eval_weighted_fpr": 0.017277593332881633, |
|
"eval_weighted_sensitivity": 0.8024786986831913, |
|
"eval_weighted_specificity": 0.9711525430264618, |
|
"step": 6430 |
|
}, |
|
{ |
|
"epoch": 10.11, |
|
"learning_rate": 3.316744427164334e-05, |
|
"loss": 0.0668, |
|
"step": 6500 |
|
}, |
|
{ |
|
"epoch": 10.89, |
|
"learning_rate": 3.1871435977190256e-05, |
|
"loss": 0.0617, |
|
"step": 7000 |
|
}, |
|
{ |
|
"epoch": 11.0, |
|
"eval_accuracy": 0.8040278853601859, |
|
"eval_f1_macro": 0.7422420279485343, |
|
"eval_f1_micro": 0.8040278853601859, |
|
"eval_f1_weighted": 0.8069943234338898, |
|
"eval_loss": 1.5223894119262695, |
|
"eval_macro_fpr": 0.017602619829831197, |
|
"eval_macro_sensitivity": 0.7512050029477755, |
|
"eval_macro_specificity": 0.9853474796083734, |
|
"eval_precision": 0.8144253833246468, |
|
"eval_precision_macro": 0.745702567690635, |
|
"eval_recall": 0.8040278853601859, |
|
"eval_recall_macro": 0.7512050029477755, |
|
"eval_runtime": 85.7004, |
|
"eval_samples_per_second": 15.064, |
|
"eval_steps_per_second": 1.89, |
|
"eval_weighted_fpr": 0.017111937774771727, |
|
"eval_weighted_sensitivity": 0.8040278853601859, |
|
"eval_weighted_specificity": 0.9761843087654142, |
|
"step": 7073 |
|
}, |
|
{ |
|
"epoch": 11.66, |
|
"learning_rate": 3.057542768273717e-05, |
|
"loss": 0.0582, |
|
"step": 7500 |
|
}, |
|
{ |
|
"epoch": 12.0, |
|
"eval_accuracy": 0.7970565453137103, |
|
"eval_f1_macro": 0.7465072775190384, |
|
"eval_f1_micro": 0.7970565453137104, |
|
"eval_f1_weighted": 0.8021748784206515, |
|
"eval_loss": 1.6427809000015259, |
|
"eval_macro_fpr": 0.01832354936918133, |
|
"eval_macro_sensitivity": 0.7654834054509102, |
|
"eval_macro_specificity": 0.9849466228972559, |
|
"eval_precision": 0.8148454054398186, |
|
"eval_precision_macro": 0.7470312881962948, |
|
"eval_recall": 0.7970565453137103, |
|
"eval_recall_macro": 0.7654834054509102, |
|
"eval_runtime": 95.4565, |
|
"eval_samples_per_second": 13.524, |
|
"eval_steps_per_second": 1.697, |
|
"eval_weighted_fpr": 0.017862012544314153, |
|
"eval_weighted_sensitivity": 0.7970565453137103, |
|
"eval_weighted_specificity": 0.9771427981451267, |
|
"step": 7716 |
|
}, |
|
{ |
|
"epoch": 12.44, |
|
"learning_rate": 2.9279419388284086e-05, |
|
"loss": 0.0511, |
|
"step": 8000 |
|
}, |
|
{ |
|
"epoch": 13.0, |
|
"eval_accuracy": 0.8195197521301317, |
|
"eval_f1_macro": 0.7586168211948496, |
|
"eval_f1_micro": 0.8195197521301317, |
|
"eval_f1_weighted": 0.8186828900418365, |
|
"eval_loss": 1.4951962232589722, |
|
"eval_macro_fpr": 0.016182326988526896, |
|
"eval_macro_sensitivity": 0.7580273333502953, |
|
"eval_macro_specificity": 0.9863597979839632, |
|
"eval_precision": 0.8208277188287908, |
|
"eval_precision_macro": 0.7645206618886077, |
|
"eval_recall": 0.8195197521301317, |
|
"eval_recall_macro": 0.7580273333502953, |
|
"eval_runtime": 85.8279, |
|
"eval_samples_per_second": 15.042, |
|
"eval_steps_per_second": 1.887, |
|
"eval_weighted_fpr": 0.01548687271518777, |
|
"eval_weighted_sensitivity": 0.8195197521301317, |
|
"eval_weighted_specificity": 0.9758772176293167, |
|
"step": 8359 |
|
}, |
|
{ |
|
"epoch": 13.22, |
|
"learning_rate": 2.7983411093831004e-05, |
|
"loss": 0.0403, |
|
"step": 8500 |
|
}, |
|
{ |
|
"epoch": 14.0, |
|
"learning_rate": 2.668740279937792e-05, |
|
"loss": 0.0476, |
|
"step": 9000 |
|
}, |
|
{ |
|
"epoch": 14.0, |
|
"eval_accuracy": 0.7970565453137103, |
|
"eval_f1_macro": 0.7416650578915932, |
|
"eval_f1_micro": 0.7970565453137104, |
|
"eval_f1_weighted": 0.7913296065268454, |
|
"eval_loss": 1.7132102251052856, |
|
"eval_macro_fpr": 0.018853564171086353, |
|
"eval_macro_sensitivity": 0.7328360987022614, |
|
"eval_macro_specificity": 0.9845211676049321, |
|
"eval_precision": 0.795801987953062, |
|
"eval_precision_macro": 0.763738385548772, |
|
"eval_recall": 0.7970565453137103, |
|
"eval_recall_macro": 0.7328360987022614, |
|
"eval_runtime": 90.1914, |
|
"eval_samples_per_second": 14.314, |
|
"eval_steps_per_second": 1.796, |
|
"eval_weighted_fpr": 0.017862012544314153, |
|
"eval_weighted_sensitivity": 0.7970565453137103, |
|
"eval_weighted_specificity": 0.9707609687602743, |
|
"step": 9002 |
|
} |
|
], |
|
"logging_steps": 500, |
|
"max_steps": 19290, |
|
"num_train_epochs": 30, |
|
"save_steps": 500, |
|
"total_flos": 2.206283653350605e+16, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|